llama-bench : log benchmark progress more verbosely

Author: akx

examples/llama-bench/llama-bench.cpp

@@ -1516,10 +1516,10 @@ int main(int argc, char ** argv) {
     llama_model * lmodel = nullptr;
     const cmd_params_instance * prev_inst = nullptr;
 
-    int params_idx = 1;
+    int params_idx = 0;
     for (const auto & inst : params_instances) {
-        LOG_TEE("llama-bench: starting benchmark %d/%ld\n", params_idx, params_instances.size());
         params_idx ++;
+        LOG_TEE("llama-bench: benchmark %d/%ld: starting\n", params_idx, params_instances.size());
         // keep the same model between tests when possible
         if (!lmodel || !prev_inst || !inst.equal_mparams(*prev_inst)) {
             if (lmodel) {
@@ -1569,10 +1569,12 @@ int main(int argc, char ** argv) {
 
         // warmup run
         if (t.n_prompt > 0) {
+            LOG_TEE("llama-bench: benchmark %d/%ld: warmup prompt run\n", params_idx, params_instances.size());
             //test_prompt(ctx, std::min(t.n_batch, std::min(t.n_prompt, 32)), 0, t.n_batch, t.n_threads);
             test_prompt(ctx, t.n_prompt, 0, t.n_batch, t.n_threads);
         }
         if (t.n_gen > 0) {
+            LOG_TEE("llama-bench: benchmark %d/%ld: warmup generation run\n", params_idx, params_instances.size());
             test_gen(ctx, 1, 0, t.n_threads);
         }
 
@@ -1582,9 +1584,11 @@ int main(int argc, char ** argv) {
             uint64_t t_start = get_time_ns();
 
             if (t.n_prompt > 0) {
+                LOG_TEE("llama-bench: benchmark %d/%ld: prompt run %d/%d\n", params_idx, params_instances.size(), i + 1, params.reps);
                 test_prompt(ctx, t.n_prompt, 0, t.n_batch, t.n_threads);
             }
             if (t.n_gen > 0) {
+                LOG_TEE("llama-bench: benchmark %d/%ld: generation run %d/%d\n", params_idx, params_instances.size(), i + 1, params.reps);
                 test_gen(ctx, t.n_gen, t.n_prompt, t.n_threads);
             }